後設分析之方法論問題之探討

Methodological Problems of meta-analysis

馬信行(Hsen-Hsing Ma)

《αβγ量化研究學刊》 1卷 (2007/12) Pp. 170-183

摘要

本文敘述組間比較設計之實證研究之後設分析方法並討論在分析調節變項平均值差異時之方法論問題。在方法上本文闡明搜尋實證研究之方法，登錄自變項；依變項；要比較之兩組之樣本數、平均值及標準差；實驗設計類型；樣本之年齡；及可轉換為效應量之t、F、x^2及r統計值。轉換公式也連同Excel計算公式一一列出。後設分析之方法論問題注重在檢定組間平均值差異顯著性時，若違反參數統計的基本假設（如殘差要均質及殘差要獨立分配）會產生何種後果。當參數統計之基本假設被違反時，本文建議使用非參數統計。

關鍵字

後設分析 ； 參數統計之基本假設 ； 效應量

並列摘要

Department of Education, National Chengchi University Abstract This article describes the method of meta-analysis of empirical studies using between groups designs and discusses methodological problems in analyzing the significance of differences between moderators. The present article delineates methods of locating studies, encoding the independent variables, dependent variables, sample sizes, means, standard deviations of comparison groups, types of experimental designs, ages of samples, and the statistical value of t, F, x^2, or r. The formulas convert statistical values to effect sizes as well as their calculating formula in Excel listed.
The discussion of methodological problems of meta-analysis was focused on the consequences of violating the assumptions of parametrical statistics, such as the assumptions of homogeniety of residuals and the independent distribution of residuals. Nonparametric statistic was suggested in case of violation of assumptions of parametric statistics.

並列關鍵字

assumptions of parametric statistics ； effect size ； meta-analysis